Unet-based improved infrared image photovoltaic panel boundary segmentation method under view angle of unmanned aerial vehicle

The invention discloses a deep learning-based photovoltaic panel semantic segmentation method applied to an infrared image. The method comprises the following steps of: establishing a photovoltaic panel data set under an unmanned aerial vehicle visual angle infrared light condition and preprocessing the photovoltaic panel data set; constructing an improved Unet semantic segmentation deep learning model; putting training sets into the improved Unet semantic segmentation deep learning model batch by batch for iteration, and testing the performance of the model obtained through real-time training through a test set; and inputting a to-be-detected photovoltaic panel image under the infrared light condition into the model corresponding to the minimum loss so as to process the to-be-detected photovoltaic panel image, and performing outputting to obtain a segmentation result. According to the method of the invention, the deep learning method is applied to the boundary detection of the infrared photovoltaic panel, and the Unet network model is improved, more significant shallow features are put forward to improve the segmentation precision of the photovoltaic panel.